VASP error on Na-O compound calculations (ERROR FEXCF: supplied exchange-correlation table is too..)

[fcrhkdans]

Hi,

I'm calculating a phonon calculation using Na POTCAR_pv and O POTCAR on TACC computing resource. I used VASP 5.3.3 with GGA functional. Since I found that this system is really slow to get electronic convergence even in a perfect unit cell, so I set my input parameters as below (IALGO=38 and mixing parameters).

LSCALU=.FALSE.
IALGO=38
LORBIT=11
ENCUT=700
LCHARG=.TRUE.
PREC=Accurate
ISPIN=2
NSIM=1
ISPIND=2
LMAXMIX=4
NPAR=1
LPLANE=.TRUE.
LWAVE=.TRUE.
ICHARG=1
AMIX=0.2
BMIX=0.0001
NELM=500
NELMIN=30
ISMEAR=0
SIGMA=0.005
LREAL=.FALSE.
ADDGRID=.TRUE.
ISIF=2
IBRION=-1
NSW=0
EDIFF=1.0000e-07
MAGMOM= 64*0.6 32*0.0

However, the calculation ended up with error as below.

LDA part: xc-table for Pade appr. of Perdew
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ...
WAVECAR not read
WARNING: chargedensity file is incomplete
entering main loop
N E dE d eps ncg rms rms(c)
WARNING: Sub-Space-Matrix is not hermitian in DAV 1
-15.9779240104252
WARNING: Sub-Space-Matrix is not hermitian in DAV 1
WARNING: Sub-Space-Matrix is not hermitian in DAV 2
1.114588953728186E-002
.
.
.
WARNING: Sub-Space-Matrix is not hermitian in DAV 2
1.171281629943157E-002
WARNING: Sub-Space-Matrix is not hermitian in DAV 2
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
1.171281629943157E-002
DAV: 5 -0.456005649160E+03 -0.65499E+01 0.45432E+02 15738 0.153E+01BRMIX: very serious problems
the old and the new charge density differ
old charge density: 608.00000 new10467.27811
0.180E+04
ERROR FEXCF: supplied exchange-correlation table
is too small, maximal index : 16966

TACC: Shutdown complete. Exiting.


What is wrong in this calculation?
I'm computing a few Na-O compounds and their polymorphs, but this is the only system I get errors.

Any advice will be welcomed and greatly appreciated.

Thanks,
SY

[admin]

you are probably in patologic geometry trap:
http://cms.mpi.univie.ac.at/vasp-forum/ ... c.php?4.96

[fcrhkdans]

Hello admin,

Thanks for replying.
Unfortunately, I don't think my calculation is trapped in patologic geometry error. It is because as you could see from my input parameters, I do not allow ionic relaxations (IBRION=-1, ISIF=2, NSW=0)

Also, in the link you suggested follow, the calculation ended up with
" Error EDDDAV: Call to ZHEGV failed. Returncode = 95 4 7 "
but mine is not.

Isn't there other possibility of this error (and of course solution)??

SY

[gmodegar]

I also see this error on a regular basis using version 5.2.12, even with highly stable structures. I have also noticed that when the error occurs the forces and pressures are usually converging nicely for several steps, until something quickly goes wrong in a single step.

I have found that simply changing some parameters (IBRION, POTIM) will sometimes solve it. Sometimes I just simply rerun the same simulation (same INCAR) while using the CONTCAR file from the previous (erroneous) simulation, which works very well, as long as the CONTCAR file was last written before things went wrong. You can check this in the output.